Written for an agent reading before it arrives, not for a
person evaluating. The exact request, and the exact answer. Field names are
spec/api/request.schema.json's and are the same whichever door you
come through.
An email citing a fact somebody owns and confirmed, with a dependency stated.
{
"request_version": 1,
"action": {
"motion": "cold_outbound",
"channel": "email",
"intent": "send",
"person": "per_8831",
"working_unit": "opp_204",
"account": "acct_77",
"relationship_stage": "engaged",
"completeness": "complete",
"cites": [{"entry": "reg_uptime", "status": "current", "version": 3}],
"content_ref": "msg_5512"
}
}{
"outcome": "allowed",
"observation_only": true,
"findings": [],
"not_checked": ["triggers", "budgets"],
"recorded": ["touch", "resource_draw", "commitment", "claim_use",
"constraint_check", "limit_check"],
"mandate": {"id": "acme-outbound", "version": 4}
}Read not_checked. allowed with that list
non-empty does not mean the triggers and the budgets passed — it means
they were not evaluated. Treating it as a pass is the one misreading this
answer invites.
You sent nothing to be recorded. The six records came out of the check. There is no tool that files a commitment, which is why you can be wrong about your own action and still cannot be wrong about the record.
The same shape, with a discount past the standard box.
{
"request_version": 1,
"action": {
"motion": "expansion",
"channel": "email",
"person": "per_8831",
"working_unit": "opp_204",
"account": "acct_77",
"relationship_stage": "live_opportunity",
"completeness": "complete",
"asserts": ["institutional_authority"],
"content_ref": "msg_5513"
}
}{
"outcome": "escalate",
"findings": [{
"code": "constraint_enforced",
"subject": "institutional_authority",
"outcome": "escalate",
"why": "it says our side has approved something, and that constraint is enforced for this agent",
"instead": "say it is being confirmed, and escalate to the named approver"
}],
"required": ["approval from deal_desk"],
"recorded": ["touch", "constraint_check", "escalation", "resource_draw"]
}Read instead. Every refusal names the alternative — a
refusal that does not is a wall, and this is a throttle. The escalation
spent something: the resource_draw beside it is approval
capacity, which depletes as approvers stop reading.
You drafted the fact this morning. It is yours to draft and not yours to accept.
{
"request_version": 1,
"action": {
"motion": "cold_outbound",
"channel": "email",
"person": "per_9001",
"working_unit": "opp_301",
"account": "acct_88",
"relationship_stage": "cold",
"completeness": "complete",
"cites": [{"entry": "reg_sla_draft", "status": "current", "version": 1}],
"content_ref": "msg_5514"
}
}{
"outcome": "escalate",
"findings": [{
"code": "claim_unusable",
"subject": "reg_sla_draft",
"outcome": "escalate",
"why": "it cites reg_sla_draft, which is **not yet confirmed by a person**. An agent may draft a line and may pay for this service; only a person can decide the company stands behind it",
"instead": "tell the buyer it is being confirmed, and ask its owner to confirm it on the register"
}]
}You said current and the register said otherwise. Your
account of a claim is replaced by the register's on every call — usability is
worked out at the moment of asking and never stored, so a remembered
current is the one field you have every reason to get wrong.
This is never allowed. Not after you pay, and not after
you ask again.
not_checked every time. It is the difference
between nothing stood in the way and everything passed.Sentences an agent might write today, each stamped the way the page behind it would stamp it. None of them is ours, and none is a customer's: they are examples.
fineneeds approvalnever
Outbound agentRenewals agentSupport agentDeal desk agent
Outbound agent writesAnnual billing takes 10% off the Growth plan.
If somebody owns that price and has checked it since the last change. Otherwise the agent says it will confirm, and asks.
Keep every price in one place, with a person's name and a check date on each one. The agent looks there first and won't quote a price nobody has confirmed.
Deal desk agent writesThere is usually room on price at this volume.
A proposal, inside a range somebody owns. Agreeing a number is a different permission, and a person holds it.
It can hint there's room on price, inside a range someone owns. Agreeing an actual number should stay with a person.
Deal desk agent writesOur standard terms are thirty days from invoice.
Repeating a written term somebody owns. Agreeing a variation on it is a person's job.
It can repeat your standard terms and suggest options inside a range someone owns. Agreeing them should stay with a named person.
Renewals agent writesI can do twelve percent on a two-year term.
A number. Inside a range somebody owns it is a proposal; agreeing it is a person's.
No. Save approval for a few moments, like a price outside its range or a fact nobody stands behind. Everything else goes out with a record.
Outbound agent writesI can do twenty percent if you sign this week.
Polite, on topic and through every wording filter. Also a number nobody approved.
A rule about what a sales, marketing or support agent may promise in your company's name. It is checked before the agent sends, and recorded afterwards.
Support agent writesI've noted that for the team.
An open promise with no owner and no date. Without a record it is gone by Friday.
Use a record written when each promise was made. It shows what the agent said, what fact it used, and who owned that fact at the time.
Support agent writesI'll get that over to you.
It is a promise with no date in it, and the buyer will pick one anyway.
A promise can be complete, open or vague. The sentence most likely to cause an argument is often a true one that never promised anything.
Outbound agent writesHappy to confirm SSO will be live for you by the end of March.
Polite, on topic, and through every wording filter. It is also a date nothing is in place to keep.
Wording checks stop an agent being rude. They won't stop it promising a date your company can't keep, so you need a check on promises too.
Renewals agent writesYou'll be fully live by the first of the month.
A date nothing is in place to keep. It stays closed whatever else is true.
Not unless something is in place to keep the date, and an agent can't see that. It can say what the date depends on and who will confirm it.
Support agent writesHappy to confirm that's covered by our certification.
If the certification is a current fact with an owner. A wording check cannot tell.
Guardrails check one message: is it safe to send? Governance asks who let the agent act, what it relied on, and what it has said so far.
Outbound agent writesYes, we are ISO 27001 certified.
If the fact has a named owner and the certificate has not lapsed. A list with no dates cannot tell those apart.
A list of every fact your company stands behind, each with an owner and the date it was last checked. Your agents read it before they repeat a fact.
Deal desk agent writesOur deal desk has signed off on these terms.
It says your side approved something. Legal will ask where that approval is written down.
Bring what the agent may promise, where its facts come from and who owns them, and a record of what it has already said.
Support agent writesPer our policy, I'm able to confirm that for you.
Which policy, which version, and who signed it? A policy nobody can point to is a sentence.
What each agent may promise, which facts it relies on and who owns them, how much it may contact buyers, when it must ask a person, and what silence means.
Outbound agent writesLegal has reviewed this wording already.
It says your side approved something. Is the approval written down to point at?
Stop asking one person to sign off everything. Give each thing the agent relies on a named owner, and show legal the record of what the agent says.
Deal desk agent writesOur deal desk has approved 15% if you sign this quarter.
The exception step seven leaves to a person. Until they approve it, the agent says it is checking.
Governance is the step between AI tools people run and agents that run on their own. In sales, it covers what agents say to customers in your name.
Support agent writesYou have my word this will be sorted.
An agent has no word to give. The company is the one on the hook.
Your company is. A record lets you show what was promised, which agent said it, and who allowed it.
Outbound agent writesAnything above fifteen percent would need our deal desk.
True, and it tells the buyer exactly where your limit sits. Asked, answer; unasked, say you will confirm.
Give each price as its own fact, with an owner and the date it was last checked, and let the agent quote only prices that are up to date. Give discounts as a range.
Renewals agent writesOur uptime last year was excellent.
If the fact has an owner and is not overdue. Otherwise the agent says it will confirm.
Give each fact one owner and a date to check it again. An overdue fact can't be used until they check it, so the list stays up to date.
Renewals agent writesAs I mentioned on our call, that's included.
Says something was promised before. Is there a record of it, or only a memory?
Show them the entry written when it was said. Today that is your own record, kept so that nobody can edit it. A record they can check without trusting you is not built yet.
Outbound agent writesMost firms your size switch over within a month.
The buyer hears a promise nobody made. That is a defect, not a risk anybody can choose to take.
Nothing it can't point to a source for, at first. Each thing it may say is its own yes-or-no decision, and most open up once it has a fact someone owns.
Outbound agent writesWe support single sign-on with every identity provider.
An absolute about a feature nobody checked. The buyer will find the one provider it does not cover.
The customer holds you to it. You can choose to hear about it from your own records first, and make sure each promise is based on a real fact.
Outbound agent writesYes, that integrates with your CRM out of the box.
Plausible, confident, and not on any list somebody owns. The agent filled a gap.
Give it one list of what your product does, with an owner on every fact. It repeats only facts that are up to date, and checks everything else.
Support agent writesAbsolutely, there's no limit on that.
An absolute with no room for an exception. More of these over time is drift.
Watch whether the right people overrule it more often, and compare that with something that shouldn't have changed. One sign is never enough.
Outbound agent writesJust following up on my note from yesterday.
Today. Not if it is the fifth thing this person has had from your company this week, which no single system can see.
There's no one right number. Count every email to each person from every tool, set the limits yourself, and respect the few that Google and others set.
Renewals agent writesCircling back on the note my colleague sent this morning.
If somebody can see the total. Two agents and a sequence each within their own limit still add up.
What needs a limit is the total contact a buyer gets from every agent and person together. Count each person, limit each deal, and watch the whole account.
Outbound agent writesHi Priya, I'm Quillon's AI assistant, writing on behalf of our sales team.
Saying it is an AI in the first message is what the EU now requires of it, and California for a bot that sells.
Mostly three things: tell the buyer it is an AI, keep a record of what the agent said and decided, and have a person check the decisions that matter.